key: cord-301556-f3m9gwvo
authors: Huang, Jessie; Hume, Adam J.; Abo, Kristine M.; Werder, Rhiannon B.; Villacorta-Martin, Carlos; Alysandratos, Konstantinos-Dionysios; Beermann, Mary Lou; Simone-Roach, Chantelle; Lindstrom-Vautrin, Jonathan; Olejnik, Judith; Suder, Ellen L.; Bullitt, Esther; Hinds, Anne; Sharma, Arjun; Bosmann, Markus; Wang, Ruobing; Hawkins, Finn; Burks, Eric J.; Saeed, Mohsan; Wilson, Andrew A.; Mühlberger, Elke; Kotton, Darrell N.
title: SARS-CoV-2 Infection of Pluripotent Stem Cell-derived Human Lung Alveolar Type 2 Cells Elicits a Rapid Epithelial-Intrinsic Inflammatory Response
date: 2020-09-18
journal: Cell Stem Cell
DOI: 10.1016/j.stem.2020.09.013
sha: 
doc_id: 301556
cord_uid: f3m9gwvo

A hallmark of severe COVID-19 pneumonia is SARS-CoV-2 infection of the facultative progenitors of lung alveoli, the alveolar epithelial type 2 cells (AT2s). However, inability to access these cells from patients, particularly at early stages of disease, limits an understanding of disease inception. Here we present an in vitro human model that simulates the initial apical infection of alveolar epithelium with SARS-CoV-2, using induced pluripotent stem cell-derived AT2s that have been adapted to air-liquid interface culture. We find a rapid transcriptomic change in infected cells, characterized by a shift to an inflammatory phenotype with upregulation of NF-kB signaling and loss of the mature alveolar program. Drug testing confirms the efficacy of remdesivir as well as TMPRSS2 protease inhibition, validating a putative mechanism used for viral entry in alveolar cells. Our model system reveals cell-intrinsic responses of a key lung target cell to SARS-CoV-2 infection and should facilitate drug development.

Responding to the COVID-19 pandemic caused by the novel coronavirus, SARS-CoV-2, requires access to human model systems that can recapitulate disease pathogenesis, identify potential targets, and enable drug testing. Access to primary human lung in vitro model systems is a particular priority since a variety of respiratory epithelial cells are the proposed targets of J o u r n a l P r e -p r o o f viral entry (Hoffmann et al., 2020; Hou et al., 2020; Zhu et al., 2020) . A rapidly emerging literature now indicates that a diversity of epithelial cells of the respiratory tract from the nasal sinuses and proximal conducting airways through the distal lung alveoli express the cell surface receptor for SARS-CoV-2, angiotensin-converting enzyme 2 (ACE2), and appear permissive to infection with SARS-CoV-2 in vivo, and in some cases in vitro (Sungnak et al., 2020; Leung et al., 2020; Ziegler et al., 2020; Sun et al., 2020; Bradley et al., 2020; Hou et al., 2020; Schaefer et al., 2020; Hou et al., 2020) . The most severe infections with SARS-CoV-2 result in acute respiratory distress syndrome (ARDS), a clinical phenotype that is thought to arise in the setting of COVID-19 pneumonia as the virus progressively targets the epithelium of the distal lung, particularly the facultative progenitors of this region, alveolar epithelial type 2 cells (AT2s) (Hou et al., 2020) . While small animal models such as Syrian hamster (Imai et al., 2020; Sia et al., 2020) and humanized ACE2 transgenic mice (Bao et al., 2020; Jiang et al., 2020) have shown changes in the alveolar epithelium after SARS-CoV-2 infection, little is known about the initial responses of human lung alveoli to SARS-CoV-2 infection due in part to the inability to access these cells from patients, particularly at early stages of disease.

To date, primary human AT2s that are harnessed from explanted lung tissues require 3D coculture with supporting fibroblasts, cannot be maintained in culture for more than 3 passages, and tend to rapidly lose their AT2 phenotype ex vivo (Jacob et al., 2019) . Thus, SARS-CoV-2 infection modeling has to this point been predominantly performed using either human airway (non-alveolar) cells in air-liquid interface cultures, non-human cell lines that naturally express the ACE2 viral receptor, such as the African Green Monkey Vero E6 cell line (Harcourt et al., 2020) , or transformed human cell lines with or without forced over-expression of ACE2.

Although some of these cell lines, such as A549 and Calu-3 cells, were originally generated from transformed cancerous lung epithelial cells, they no longer express NKX2-1, the master J o u r n a l P r e -p r o o f transcriptional regulator required for differentiated lung epithelial gene expression (Abo et al., 2020) , and thus are limited in their capacity to simulate an accurate lung cellular response to most perturbations, including viral infections.

To provide alternative sources of self-renewing human lung epithelial lineages, our group and others have recently developed human lung epithelial organoids and 2D air-liquid interface (ALI) lung cultures through the directed differentiation of induced pluripotent stem cells (iPSCs) in vitro (Abo et al., 2020; Hawkins et al., 2017; Huang et al., 2014; Hurley et al., 2020; Jacob et al., 2017; Longmire et al., 2012; McCauley et al., 2018a; McCauley et al., 2018b; McCauley et al., 2017; Serra et al., 2017; Yamamoto et al., 2017) . Here we report the successful infection of a pure population of human iPSC-derived AT2-like cells (iAT2s) with SARS-CoV-2, providing a reductionist model that reveals the cell-intrinsic distal lung epithelial global transcriptomic responses to infection. By 1 day post-infection (dpi), SARS-CoV-2 induced a rapid global transcriptomic change in infected iAT2s characterized by a shift to an inflammatory phenotype associated with the secretion of cytokines encoded by NF-kB target genes. By 4 dpi, there were time-dependent epithelial interferon responses and progressive loss of the mature lung alveolar epithelial program, exemplified by significant downregulation of surfactant encoding genestranscriptomic changes that were not predicted by prior human airway or cell line models. Our model system thus reveals the cell-intrinsic responses of a key lung target cell to infection, providing a novel, physiologically-relevant platform for further drug development and facilitating a deeper understanding of COVID-19 pathogenesis.

In order to develop a human model system, we used the technique of directed differentiation Jacob et al., 2019) to generate iAT2s from either human embryonic stem cells or iPSCs engineered to carry a tdTomato reporter targeted to the endogenous SFTPC locus (Hurley et al., 2020; Jacob et al., 2017) . In 3D Matrigel cultures, we established selfrenewing epithelial spheres composed of purified iAT2s, >90% of which expressed surfactant protein-C (SFTPC), the canonical AT2 marker, as monitored by flow cytometry assessment of the SFTPC tdTomato reporter at each passage in culture ( Figure 1A , B). Serially passaging these epithelial spheres generated >10 30 iAT2s per starting sorted tdTomato+ cell over 225 days in culture (Hurley et al., 2020) , generating cells that maintained expression of AT2 marker genes including surfactants as shown by single cell RNA sequencing (scRNA-Seq) ( Figure 1C -D, Figure S1 ) and providing a stable source of human primary-like AT2 cells for viral infection disease modeling.

Since the directed differentiation of human iPSCs in vitro is designed to recapitulate the sequence of developmental milestones that accompanies in vivo human fetal organogenesis, we first sought to understand the ontogeny of expression of the coronavirus receptor, ACE2, during human fetal development from early endodermal progenitors through increasingly more mature stages of alveolar development. By analyzing our previously published transcriptomic time series profiles of developing human fetal and adult primary AT2s (Hurley et al., 2020) , we found that ACE2 expression increases from early to late canalicular stages of distal human lung development, with expression levels similar to adult AT2 levels present by week 21 of gestation J o u r n a l P r e -p r o o f in developing alveolar epithelial cells ( Figure 1E ). We found the directed differentiation of human pluripotent stem cells (RUES2 embryonic stem cells and SPC2 iPSCs) in vitro into purified distal lung AT2-like cells resulted in cells expressing similar levels of ACE2 to adult primary cell controls in head to head comparisons ( Figure 1E and Figure S1 , respectively). We and others (Abo et al., 2020; Ziegler et al., 2020) have recently profiled the frequency of ACE2 mRNA expressing primary adult and iPSC-derived AT2-like cells by scRNA-Seq ( Figure S1 ), finding that mRNA expression occurs in a minority of cells (1-3%) at any given time with similar frequencies observed in primary AT2s compared to iAT2s. In contrast, the gene encoding the protease utilized for viral entry, TMPRSS2, is expressed more robustly in both AT2s and iAT2s ( Figure S1 and (Abo et al., 2020) ) and is less developmentally variable, being stably expressed by week 16 of distal fetal lung development ( Figure 1E ).

Because lung epithelial infection by SARS-CoV-2 occurs at the apical membrane of cells facing the air-filled lumens of airways and alveoli, submerged cultures of 3D epithelial spheres with apical membranes oriented interiorly are unlikely to faithfully recapitulate infection physiology.

Therefore, we adapted our SFTPC tdTomato+ iAT2s (SPC2-ST-B2 iPSC line) to 2D ALI cultures ( Figure   1F ), generating monolayered epithelial cultures of pure iAT2s with apical-basal polarization and barrier integrity (transepithelial electrical resistance = 454 ± 73 ohms x cm 2 ), while preserving or augmenting expression of AT2-specific genes (e.g. SFTPC, SFTPB, SFTPA1, SFTPA2, SFTPD, NAPSA, and PGC) as detailed in our recent preprint (Abo et al., 2020) as well as Figure S1.

To quantify protein-level expression frequency of ACE2 in iAT2s in ALI cultures, we employed flow cytometry, observing that 14.7 ± 6.6% of live iAT2s demonstrated cell surface expression of J o u r n a l P r e -p r o o f ACE2 ( Figure 1G ) and indicating more frequent expression at the protein level than had been predicted from analysis of published primary AT2 or ALI-cultured iAT2 scRNA-Seq profiles ( Figure S1 and (Abo et al., 2020) ). We validated the sensitivity and specificity of the ACE2 antibody by staining controls consisting of human 293T cells, which lack ACE2, and 293T cells lentivirally transduced to over-express ACE2 (293T-ACE2; Figure 1G ) (Crawford et al., 2020) .

Apical localization of ACE2 protein in iAT2s was confirmed by immunofluorescence staining ( Figure 1H ). To test functionality of ACE2 as the viral receptor, we used a GFP-expressing lentivirus pseudotyped with either viral spike protein (S) or VSV-G envelope to infect iAT2s, 293T cells, and 293T-ACE2 cells, finding that both pseudotypes infected iAT2s and 293T-ACE2 cells whereas the Spike pseudotype did not infect 293T cells ( Figure 1I , J; Figure S1C ).

To test whether ALI cultures of iAT2s are permissive to SARS-CoV-2 infection, we employed escalating viral doses representing a broad range of multiplicities of infections (MOIs). iAT2s in ALI culture were apically exposed to SARS-CoV-2 for 1 hour (Figure 2A Figure 2H ), and iAT2s could be infected while growing either in ALI cultures or when grown as 3D spheres that were mechanically disrupted at the time of infection ( Figure 2I ). In iAT2 ALI J o u r n a l P r e -p r o o f cultures, infectious virus was predominantly released from the apical side at increasing titers over time (4 dpi vs 1 dpi; Figure 2J ), providing further evidence of ongoing viral replication in iAT2s in this model system. While initial viral infection efficiency was similar to the efficiency of ACE2+ protein expressing iAT2s, the discrepancy between the number of ACE2+ iAT2s (13%) and N+ cells at 4 dpi (60%) may suggest alternative mechanisms of viral entry or spread. It is also conceivable that ACE2 surface expression of some SARS-CoV-2 target cells was below our flow cytometry-based detection limit. Immunofluorescence analysis of infected iAT2s 

Having established a putative human model system for SARS-CoV-2 infection of AT2-like cells, we next sought to define the global, time-dependent transcriptomic responses of cells to infection. We performed bulk RNA sequencing of SARS-CoV-2-infected iAT2s at 1 dpi or 4 dpi, compared to mock-infected iAT2 controls (n=3 biological replicates at each time point, Figure   3A ). Profound and progressive changes in the global transcriptomes of infected iAT2s were observed based on principle components analyses and differential gene expression ( Figure 3B , Table S1 ). For example, 4519 genes were differentially expressed between mock and SARS-CoV-2-infected cells at 1 dpi, 10725 genes between SARS-CoV-2-infected cells at 1 dpi and 4 J o u r n a l P r e -p r o o f dpi, and 10462 between the infected samples as a whole and mock (FDR<0.05; empirical Bayes moderated t-test; Table S1 ). Viral transcripts, including the viral genome, were amongst the top differentially expressed transcripts at both 1 dpi and 4 dpi, representing 28% to 33% of all reads mapped at 1 dpi ( Figure 3C ). AT2-specific genes, such as SFTPC, were amongst the top downregulated genes by 1 dpi, and progressive loss of the AT2 program continued to 4 dpi with significant continued loss of SFTPC, SFTPA1, SFTPD, and SFTPC tdTomato encoding transcripts (FDR<0.05; Table S1; Figure 3C ). This loss of AT2 marker gene expression was not accompanied by any detectable emergence of alternate lung fates as there was no transcriptomic evidence of any upregulation of airway (SCGB1A1, FOXJ1, FOXI1, TP63, MUC5B, or MUC5AC) or alveolar type 1 cell (AGER; Fig. 3C ) markers. Gene set enrichment analysis (GSEA) revealed significant upregulation of inflammatory pathways both at 1 dpi (FDR<0.05; Figure 3D ) and 4 dpi, with NF-kB mediated inflammatory signaling representing the first-and second-most upregulated pathways at 1 dpi and 4 dpi, respectively, and the most upregulated pathway over the entire time course ( Figure 3D -G). Interferon (IFN) signaling was in the top 10 pathways significantly upregulated at both time points, and pathways reported to be activated by interferon signaling were also significantly upregulated (FDR<0.5) including KRAS (MAPK) signaling, IL6-JAK-STAT3, and IL2-STAT5 signaling ( Figure 3D ). Despite the GSEA findings, the global transcriptomic analysis showed no significant induction of individual type I and III IFN genes (e.g. IFNB1, IFNL1, or IFNL2) at any time point post-infection, and upregulation of multiple IFN signaling related genes and targets (IFNAR2, IRF 1/4/7/8/9, IFIT1, MX1, CXCL10, CXCL11, SOCS3, and ISG15) was observed mainly at 4 dpi rather than 1 dpi, consistent with delayed signaling ( Figure 3F -H). These results were confirmed by RT-qPCR ( Figure 4 , Figure S4 ) and indicate that SARS-CoV-2 infection of iAT2s elicits a delayed, modest IFN response compared to treatment of iAT2s with either IFNβ or transfection with the synthetic double-stranded RNA analog, poly(I:C), which we found stimulate stronger interferon responses ( Figure S4 ). Compared to other published datasets of SARS-CoV-2 infection models in Calu-3, J o u r n a l P r e -p r o o f A549-ACE2, and primary (non-alveolar) normal human bronchial epithelium (NHBE) (Blanco-Melo et al., 2020) , iAT2s were able to uniquely capture the downregulation of AT2-specific programs, such as decreased surfactant gene expression and loss of lamellar body gene ontology (GO) terms (comparative gene set enrichment based on lung-related GO biological processes; Figure S3 ).

Consistent with the cytopathogenicity suggested by our microscopy studies, cell death-related genes were significantly upregulated over the entire time course post-infection, and stressrelated signaling was evident by 4 dpi as multiple heat shock proteins were in the top most upregulated transcripts, comparing 4 dpi to 1 dpi ( Figure 3H ; e.g. HSPA1A, HSPA1B, HSPA6, HSP90AB1). Significant downregulation of proliferation markers (TOP2A and MKI67; Figure 3G , H) was evident by 4 dpi, and there was a significant decrease in cell viability ( Figure S1 ). Taken together, these results suggest that SARS-CoV-2 infection of human iAT2s results in a cellintrinsic shift away from an AT2 program toward an inflammatory program with NF-kB mediated inflammatory responses significantly upregulated and limited induction of interferon signaling.

To compare our findings to changes in AT2s in vivo, we performed pro-SFTPC immunostaining of lung tissue sections from the autopsies of two individuals who died from SARS-CoV-2 induced respiratory failure (clinical information provided in STAR Methods). In contrast to the typical, frequent pattern of pro-SFTPC immunostaining in control lung sections, COVID-19 decedent lungs exhibited regions of reduced and sporadic pro-SFTPC staining interspersed with regions of AT2 cell hyperplasia, evident in our samples as rows of cuboidal pro-SFTPC+ alveolar cells, with additional histopathologic findings of diffuse alveolar damage, such as hyaline membrane formation, as has been described in prior COVID-19 patient autopsies (Chen J o u r n a l P r e -p r o o f et al., 2020). Sloughing of cells that stained positively for cytokeratin AE1/AE3 further confirmed regional injury to the alveolar epithelium, consistent with injury observed in the iAT2 in vitro model. (Figure 4A -C).

We validated downregulation of iAT2-specific genes by RT-qPCR and observed significantly diminished expression of SFTPC, SFTPA1 and LAMP3 at 4 dpi ( Figure 4D ). Moreover, we demonstrated functional activation of NF-kB signaling in infected iAT2s, as predicted by our bioinformatics analysis, by quantifying expression of NF-kB modulated target mRNAs and proteins. Upregulation of NF-kB target transcripts IL6, CXCL8, CXCL2, CXCL3, CXCL10, and CXCL11, as well as NF-kB related mRNA NFKB1, NFKB2, RELA, and RELB, was validated by RT-qPCR ( Figure 4E ). Secretion of NF-kB target proteins by infected iAT2s was determined by Luminex analysis of apical washes and basolateral media. IL-6 and CXCL8 (IL-8) were increased both apically and basolaterally, while GM-CSF and VEGF were secreted into the basolateral media, as has been shown previously in other models of AT2 injury (Pham et al., 2002 ) ( Figure 4F ).

Finally, to assess the potential of iAT2s to screen for COVID-19 therapeutics that might target the alveolar epithelium, we tested the effect of a TMPRSS2 inhibitor, camostat mesylate, that was recently shown to block SARS-CoV-2 infection in Vero cells, Calu-3 cells, and human airway epithelial cells (Hoffmann et al., 2020) , but has not been tested previously in human alveolar cells. Camostat significantly reduced the levels of detectable viral N transcript at 2 dpi ( Figure 4H ), indicating its potential as an antiviral drug and suggesting that SARS-CoV-2 infection of iAT2s relies on priming by the protease TMPRSS2, which is expressed in both iAT2s and primary adult human AT2s (Figure 1 and S1). Conversely, the cathepsin B/L inhibitor J o u r n a l P r e -p r o o f E-64d, which blocks SARS-CoV-2 infection of Vero, HeLa-ACE2, Calu-3, and 293T-ACE2 cells ( Figure S4C ; Hoffmann et al., 2020; Ou et al., 2020; Shang et al., 2020) , had no effect in iAT2s ( Figure 4H ), suggesting that SARS-CoV-2 infection in the alveolar epithelium is independent of the endosomal cysteine proteases cathepsin B and L, even though they are expressed in both iAT2s and primary AT2s ( Figure S1 ). In addition, the FDA-approved broad-spectrum antiviral drug remdesivir (GS-5734) that is used to treat COVID-19 patients and was shown to inhibit SARS-CoV-2 in cell culture and mouse models (Pruijssers et al., 2020; Wang et al., 2020a) , significantly reduced viral N transcript ( Figure 4I ). Altogether, these data highlight the importance of using a physiologically relevant cell model to study SARS-CoV-2.

Taken together, our approach provides a new human model of SARS-CoV-2 infection of AT2s, a key lung target cell which is otherwise difficult to access in vivo and hard to maintain in vitro.

Since iAT2s can be propagated indefinitely in 3D culture in a form that is easily shareable between labs (Hurley et al., 2020; Jacob et al., 2019) , our adaptation of these cells to 2D ALI cultures now allows straight-forward simulations of apical viral respiratory infections of a selfrenewing cell type that can be scaled up nearly inexhaustibly and studied in a highly pure form, thus simulating cell-autonomous or "epithelial-only" host responses to pathogens. Our results implicate AT2s as inflammatory signaling centers that respond to SARS-CoV-2 infection within 24 hours with NF-kB signaling predominating this response, findings that are independently replicated in freshly purified adult human primary AT2 cells infected with SARS-CoV-2, as reported by Mulay et al. (2020) . The significant loss of surfactant gene expression and cellular stress, toxicity, and death of iAT2s observed in our model are likely to be clinically relevant as similar observations were evident in vivo in the lung autopsies of multiple COVID-19 decedents.

Others have previously demonstrated that primary adult human AT2s are infectible with SARS-CoV (SARS1) in vitro (Qian et al., 2013) and more recently have shown that AT2s are either infected in vivo with SARS-COV-2 in COVID-19 autopsy lungs (Bradley et al., 2020; Hou et al., 2020; Schaefer et al., 2020) or can contribute to lung regeneration in COVID-19 ARDS survivors , further highlighting the relevance of AT2s and their study in SARS-CoV-2 infection.

Importantly, IFN responses were found to be moderate in our model with only a subset of ISGs and no canonical type 1 or 3 IFN ligands (IFNB1, L1, or L2) being significantly differentially expressed in SARS-CoV-2-infected samples compared to noninfected cells. SARS-CoV-2 has been shown to be sensitive to IFNλ and IFNβ treatment, so the absence of a robust IFN response in AT2s, if verified in vivo, would have significant clinical implications and might suggest pathways to augment therapeutically in COVID-19 patients before they progress to ARDS (Blanco-Melo et al., 2020; Broggi et al., 2020; Clementi et al., 2020) . Indeed, some clinical studies suggest a beneficial effect of early type I IFN treatment on COVID-19 progression (Wang et al., 2020b) . In prior studies of cell lines that lacked AT2-specific gene expression, the absence of a robust IFN response either was overcome by using a high MOI for infection or IFN induction occurred at late time points post infection when using a low MOI . This is in contrast to our results which indicate a moderate and delayed (4 dpi) IFN response at low and high MOIs (MOI 0.4 and MOI 5). Interestingly, a delayed IFN response was also observed in SARS-CoV-and MERS-CoV-infected human airway epithelial cells and is a determinant of SARS disease severity (Channappanavar et al., 2016; Menachery et al., 2014) .

Recent studies have shown that host cell factors, such as ACE2, TMPRSS2, cathepsins, and furin are key to SARS-CoV-2 entry, and thus may be potential therapeutic targets (Hoffman et al., 2020; Ou et al., 2020) . We found that TMPRSS2 is used for viral entry in our system whereas cathepsins are not. In the iAT2 model, ACE2 is heterogeneously expressed (Figure 1) , concordant with the level of heterogeneity in adult human AT2s in vivo (Abo et al., 2020) . While the frequency of ACE2 expression in our model system at the time of infection is comparable to the initial frequency of infection of iAT2s cells, this does not necessarily prove that initial iAT2 infection is ACE2-dependent, which requires further study potentially using ACE2 knockout iPSC lines.

An important caveat of our study is the well-published observation that most human lineages derived in vitro from iPSCs are immature or fetal in phenotype, possibly confounding disease modeling. In addition, the robust self-renewing capabilities of iAT2s also distinguish them from primary AT2s that proliferate poorly ex vivo, suggesting potential differences in how they may respond to SARS-CoV-2. However, our iAT2s show expression of maturation genes, including surfactant proteins ( Figure 1D , (Hurley et al., 2020) ). Furthermore, iAT2s after adaptation to ALI culture augment expression of AT2 maturation markers and are less proliferative than those in 3D culture, and are thus more similar to primary AT2s ( Figure S1 ; Abo et al., 2020) . This, together with the observation of SARS-CoV-2 virions intracellularly in lamellar bodies and extracellularly in the vicinity of tubular myelin, confirms that surfactant-secreting, functionally mature AT2-like cells were the targets of infection in our studies. The presence of virions within lamellar bodies also implies that this surfactant-packaging organelle, specific to mature AT2 cells within the lung epithelium and absent in lung cell lines, may be a site directly utilized for and potentially dysregulated by SARS-CoV-2 infection. Thus, our model system reveals the cellintrinsic responses of a key lung target cell to infection, facilitating a deeper understanding of COVID-19 pathogenesis and providing a platform for drug discovery.

A limitation of our study is the observation that iAT2s in our culture conditions do not generate alveolar type 1 (AT1)-like cells, an ability that is similarly lacking in other published reports of in vitro cultures of both iPSC-derived and primary human AT2s to date (Barkauskas et al., 2013) .

While the basis for this divergence from mouse AT2 cell culture behavior remains unclear, it likely either reflects differences between human and mouse alveolar epithelial systems or missing factors in human in vitro culture conditions that are present in vivo in the AT2 niche, shortcomings that could potentially be addressed with future cell culture modifications.

Regardless, our findings should not detract from published observations that SARS-CoV-2 infection has also been observed in AT1 cells in COVID-19 autopsy specimens (Bradley et al., 2020) in addition to several airway epithelial cell types. Additionally, because we do not have mesenchymal support cells present in our cell cultures, we ensure that epithelial-intrinsic responses can be studied without any confounding effects that can arise in culture systems that depend on mesenchymal cell co-cultures. Although there are differences between our in vitro model and other ex vivo or in vivo models, human in vitro models recapitulating infections of those lineages or introducing further complexity by adding immune lineages are likely to similarly advance our understanding of COVID-19 disease pathogenesis in concert with our iAT2 cell model. Figure S1 ), which is apically localized, as observed by immunofluorescence staining (scale bar=10 µm) (H). (I-J) iAT2s infected with a GFP-expressing lentivirus pseudotyped with either VSVG or SARS-CoV-2 Spike envelopes, n=3. *p<0.05, oneway ANOVA, all bars represent mean +/-standard deviation. See also Figure S1 . , Hurley et al., 2020 . The human embryonic stem cell line RUES2 was a kind gift from Dr. Ali H. Brivanlou of

The Rockefeller University and we previously targeted this line with a SFTPC-tdTomato reporter . All PSC lines used in this study displayed a normal karyotype when analysed by G-banding (Cell Line Genetics). All PSC lines were maintained in feeder-free conditions, on growth factor reduced Matrigel (Corning) in 6-well tissue culture dishes (Corning), in mTeSR1 medium (StemCell Technologies) using gentle cell dissociation reagent for passaging. Further details of iPSC derivation, characterization, and culture are available for free download at http://www.bu.edu/dbin/stemcells/protocols.php.

This study was reviewed by the IRB of Boston University and found not to constitute human subjects research. With consent from next-of-kin, human lung tissues from decedents (male, age 91 and male, age 72) with COVID-19 were collected at the time of autopsies performed at Boston Medical Center, fixed in formalin and embedded.

Human iPSC lines, clones SPC-ST-B2 and BU3 NGST, underwent directed differentiation to generate iPSC-derived alveolar epithelial type II like cells (iAT2s) in 3D Matrigel cultures using methods we have previously published (Jacob et al., 2019) . As previously described (Jacob et ("alveolospheres") by plating in Matrigel (Corning) droplets at a density of 400 cells/µl with refeeding every other day in CK+DCI medium according to our previously published protocol (Jacob et al., 2019) . iAT2 culture quality and purity was monitored at each passage by flow J o u r n a l P r e -p r o o f cytometry, with >80% of cells expressing SFTPC tdTomato over time, as we have previously detailed (Jacob et al., 2019 , Hurley et al., 2020 .

To establish air-liquid interface (ALI) cultures, single cell suspensions of iAT2s were prepared as we have recently detailed (Abo et al., 2020) . Briefly, Matrigel droplets containing iAT2s as 3D

sphere cultures were dissolved in 2 mg/ml dispase (Sigma) and alveolospheres were dissociated in 0.05% trypsin (Gibco) to generate a single-cell suspension. 6.5mm Transwell inserts (Corning) were coated with dilute Matrigel (Corning) according to the manufacturer's instructions. Single-cell iAT2s were plated on Transwells at a density of 520,000 live cells/cm 2 in 100µl of CK+DCI with 10µM Rho-associated kinase inhibitor ("Y"; Sigma Y-27632). 600µl of CK+DCI+Y was added to the basolateral compartment. 24 hours after plating, basolateral media was refreshed to CK+DCI+Y. 48 hours after plating, apical media was aspirated to initiate airliquid interface culture. 72 hours after plating, basolateral media was changed to CK+DCI to remove the rho-associated kinase inhibitor. Basolateral media was changed 3 times per week thereafter. 

iAT2s plated in ALI culture were infected with purified SARS-CoV-2 stock at the indicated multiplicity of infection (MOI). 100 µl inoculum was prepared in CK+DCI media (or mock-infected with medium-only). Inoculum was added to the apical chamber of each Transwell and incubated for 1 hour at 37 o C and 5% CO2. After the adsorption period, the inoculum was removed and cells were incubated at 37 o C for 1 or 4 days. At the time of harvest, basolateral media was collected for further analysis and apical washes were performed by adding 100 µl CK+DCI media to the apical chamber, incubated for 15 minutes at room temperature before collection for further analysis. Both the apical washes and basolateral media were used for viral titration and Luminex assays as described below. Infected and mock-infected iAT2s were fixed in 10% formalin and used for immunofluorescence analysis or electron microscopy as described below.

For flow cytometry, infected and mock-infected iAT2 cells were first detached by adding 0.2 mL Accutase (A6964, Sigma) apically and incubated at room temperature for 15 minutes. Detached cells were pelleted by low speed centrifugation, resuspended in 10% formalin, and used for flow cytometry as described below. Cells were lysed in TRIzol for RNAseq and RT-qPCR analysis.

To infect iAT2 alveolospheres in 3D culture, cells were dissolved in 2 mg/ml dispase (Sigma) for 1 hour at 37°C, and alveolospheres were mechanically dissociated. Cells were washed with PBS and resuspended in inoculum (MOI 0.4) for 1 hr at 37°C. Cells were subsequently replated in matrigel droplets and collected at 2 dpi in TRIzol for RT-qPCR analysis.

For determining cell viability, infected and mock-infected iAT2s at ALI were first detached by adding 0.2 mL Accutase apically and incubated at room temperature for 15 minutes. Detached cells were pelleted by low speed centrifugation, resuspended in PBS, diluted 1:1 in trypan blue, and counted using a LUNA-II™ Automated Cell Counter (Logos Biosystems).

For nucleoprotein immunofluorescence, infected or control iAT2s on Transwell inserts were fixed in 10% formalin for 6 hours, washed twice in PBS (10 min, room temperature), permeabilized with PBS containing 0.25% Triton X-100 and 2.5% normal donkey serum (30 min, room temperature), and blocked with PBS containing 2.5% normal donkey serum (20 min, room temperature). Subsequently, cells were incubated with primary antibody diluted in anti- For ACE2 immunofluorescence, never-infected iAT2s at ALI were fixed for 10 min at RT in 4%

PFA. After fixation, the same protocol was followed as above, using an anti-ACE2 primary antibody (R&D, AF933, 1:100) or pre-immune serum for overnight incubation, and an appropriate secondary (AlexaFluor 647 Donkey Anti-Goat IgG (H+L), 1:500, Invitrogen A21447).

Pseudotyped particles carrying the pHAGE-EF1αL-GFP lentiviral vector were packaged using a 5-plasmid transfection protocol (27). In brief, 293T cells were transfected using Trans-IT (Mirus) with a plasmid carrying the lentiviral backbone (pHAGE-EF1αL-GFP; plasmid map downloadable from www.kottonlab.com), 3 helper plasmids encoding Rev, tat, and gag/pol genes, in addition to plasmids encoding either the VSV-G or the SARS-CoV-2 Spike envelope (plasmid HDM-IDTSpike-FixK, cloned by Jesse Bloom and a kind gift from Alex Balazs, (Crawford et al., 2020) 

For post infection flow cytometry, fixed iAT2s were either stained for cell surface expression of ACE2 (R&D, #AF933, 4-8µg/2.5x10 6 cells) followed by donkey anti-goat IgG-AF647 (Invitrogen, #A21447) or were permeabilized with saponin buffer (Biolegend) then stained with SARS-CoV nucleoprotein (N) antibody (Rockland, #200-401-A50, 1:1000), followed by donkey anti-rabbit IgG-AF488 (Jackson ImmunoResearch, #711-545-152). Gating was based on either mock infected stained controls or infected, isotype-stained controls. Flow cytometry staining was quantified using a Stratedigm S1000EXI and analysed with FlowJo v10.6.2 (FlowJo, Tree Star Inc). FACS plots shown represent single-cells based on forward-scatter/side-scatter gating.

iAT2 cells were collected in Qiazol (Qiagen) or TRIzol (ThermoFisher) then RNA was extracted using the RNAeasy mini kit (Qiagen) or following the manufacturer's protocol, 

Following pre-treatment, all apical media were aspirated and SARS-CoV-2 (MOI 0.04) was added for 1 hour without any drugs apically, after which the inoculum was removed. iAT2s were exposed to the compounds basolaterally for the entire duration of the experiment. Cells were harvested in TRIzol after 2 dpi and processed for RT-qPCR.

Vero E6 cells were treated with the indicated concentrations of camostat mesylate, E-64d, remdesivir, or DMSO control for 30 min at 37°C. After 30 min, SARS-CoV-2 (MOI 0.1) was added. Cells were harvested in TRIzol after 2 dpi and processed for RT-qPCR.

For immune stimulation treatments, iAT2s at ALI were treated with poly(I:C) delivered with

Oligofectamine Transfection Reagent (Invitrogen) or treated with recombinant human IFNβ (rhIFNβ). Prior to treatment, 10 µL poly(I:C) was mixed with 10 µL Oligofectamine and incubated at RT for 15 minutes. After the incubation period, the poly(I:C) and Oligofectamine mixture was added to 80 µL of CK+DCI media, for a total of 100 µL per well. iAT2s at ALI were treated apically (100 µL) with poly(I:C) (10 µg/mL) and Oligofectamine, or apically and basolaterally (600 µL) with IFNβ (10 ng/mL) for 24 hours at 37°C. Cells were subsequently harvested and processed for RT-qPCR.

Apical washes and basolateral media samples were clarified by centrifugation and analyzed 

For bulk RNA sequencing (RNA-Seq), biological triplicate (n=3) samples of purified RNA extracts were harvested from each group of samples prepared as follows. After 208 days of total time in culture, iAT2s cultured as serially passaged 3D spheres were single-cell passaged onto Transwell inserts. Apical media was removed on day 210 to initiate air-liquid interface (ALI) culture. On day 218, 6 replicate wells of iAT2s were exposed to SARS-CoV-2 in an apical inoculum and 3 replicate wells were exposed to mock. On day 219, 3 mock and 3 post-infection samples (1 dpi) were collected. Three additional post-infection samples (4 dpi) were collected on day 222. mRNA was isolated from each sample using magnetic bead-based poly(A) selection, followed by synthesis of cDNA fragments. The products were end-paired and PCRamplified to create each final cDNA library. Sequencing of pooled libraries was done using a NextSeq 500 (Illumina). The quality of the raw sequencing data was assessed using FastQC v.0.11.7. Sequence reads were aligned to a combination of the human and SARS-CoV-2 genome reference (GRCh38 and Wuhan Hu-1 isolate) and the TdTomato reporter sequence, empirical Bayes moderation was used to test differential expression (moderated t-tests). Pvalues were adjusted for multiple testing using Benjamini-Hochberg correction (false discovery rate-adjusted p-value; FDR). Differentially expressed genes between the groups in each experiment were visualized using Glimma v1.11.1, and FDR<0.05 was set as the threshold for determining significant differential gene expression. Gene set analysis was performed with

Hallmark gene sets using the Camera package (Wu and Smyth, 2012) . For the comparisons between different infection models ( Figure S3 ), we put our analyses above in context with other publicly available datasets (Riva et al., 2020 , Blanco-Melo et al., 2020 . Specifically, for the SARS-CoV-2 infections (versus mock-treated samples) of normal human bronchial epithelial cells (NHBE), A549 with forced ACE2 over-expression (A549-ACE2) and Calu-3, we used Series 1, Series 16 and Series 7 respectively from the GEO dataset GSE147507 mapped to the human genome reference (GRCh38). For the Vero E6 infection model, we used the samples from GSE153940 mapped to the African Green Monkey (Chlorecebus sabaeus) genome reference (Ensembl, ChlSab1.1). For these contrasts, log-fold-change ranked gene lists were generated using the same procedure described above. We then used FGSEA (v.1.9.7, (Korotkevich et al., 2019) ) to test for enrichment in lung-related Gene Ontology sets in the preranked gene lists. The inclusion criteria for gene sets to test was: any GO in the Molecular Signature Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb) with any of the following terms in its name: "LUNG", "SURFACTANT", "ALVEO", "LAMEL", "MULTIVESICULAR". Normalized enrichment scores (NES) for tests that resulted in FDR < 0.2 were then plotted in a heatmap ( Figure S3 ). Figure 3 . Listed are names and statistics (fold-change and falsediscovery rate-adjusted values; FDR) for all genes tested through bulk RNA-seq analysis of iAT2s infected with SARS-CoV-2 (MOI 5) at 1 and 4 dpi, along with mock-infected cells at 1 day.

Biostatistical comparisons were performed between 1 dpi and mock-infected, 4 dpi and 1 dpi, and between all infected samples vs. mock-infected samples, with an expression cut-off for significant differential expression of FDR<0.05.

J o u r n a l P r e -p r o o f

Human iPSC-derived alveolar and airway epithelial cells can be cultured at air-liquid interface and express SARS-CoV-2 host factors. bioRxiv

The pathogenicity of SARS-CoV-2 in hACE2 transgenic mice

Type 2 alveolar cells are stem cells in adult lung

Drives Development of COVID-19

Histopathology and ultrastructural findings of fatal COVID-19 infections in Washington State: a case series

Dysregulated Type I Interferon and Inflammatory Monocyte-Macrophage Responses Cause Lethal Pneumonia in SARS-CoV-Infected Mice

Pulmonary alveolar regeneration in adult COVID

Protocol and Reagents for Pseudotyping Lentiviral Particles with SARS-CoV-2 Spike Protein for Neutralization Assays

Single-cell RNA sequencing reveals profibrotic roles of distinct epithelial and mesenchymal lineages in pulmonary fibrosis

Severe Acute Respiratory Syndrome Coronavirus 2 from Patient with Coronavirus Disease, United States

Prospective isolation of NKX2-1-expressing human lung progenitors derived from pluripotent stem cells

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

SARS-CoV-2 Reverse Genetics Reveals a

Efficient generation of lung and airway epithelial cells from human pluripotent stem cells

Reconstructed Single-Cell Fate Trajectories Define Lineage Plasticity Windows during Differentiation of Human PSC-Derived Distal Lung Progenitors

Syrian hamsters as a small animal model for SARS-CoV-2 infection and countermeasure development

Differentiation of Human Pluripotent Stem Cells into Functional Lung Alveolar Epithelial Cells

Derivation of self-renewing lung alveolar epithelial type II cells from human pluripotent stem cells

Pathogenesis of SARS-CoV-2 in Transgenic Mice Expressing Human Angiotensin-Converting Enzyme 2

Activation and evasion of type I interferon responses by SARS-CoV-2

ACE-2 Expression in the Small Airway Epithelia of Smokers and COPD Patients: Implications for COVID-19

Efficient derivation of purified lung and thyroid progenitors from embryonic stem cells

Single-Cell Transcriptomic Profiling of Pluripotent Stem Cell-Derived SCGB3A2+ Airway Epithelium

Derivation of Epithelial-Only Airway Organoids from Human Pluripotent Stem Cells

Efficient Derivation of Functional Human Airway Epithelium from Pluripotent Stem Cells via Temporal Regulation of Wnt Signaling

Pathogenic influenza viruses and coronaviruses utilize similar and contrasting approaches to control interferon-stimulated gene responses

SARS-CoV-2 infection of primary human lung epithelium for COVID-19 modeling and drug discovery

Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its immune crossreactivity with SARS-CoV

Hypoxia upregulates VEGF expression in alveolar epithelial cells in vitro and in vivo

Remdesivir Inhibits SARS-CoV-2 in Human Lung Cells and Chimeric SARS-CoV Expressing the SARS-CoV-2 RNA Polymerase in Mice

Innate immune response of human alveolar type II cells infected with severe acute respiratory syndrome-coronavirus

In situ detection of SARS-CoV-2 in lungs and airways of patients with COVID-19

Pluripotent stem cell differentiation reveals distinct developmental pathways regulating lung-versus thyroid-lineage specification

Cell entry mechanisms of SARS-CoV-2

Pathogenesis and transmission of SARS-CoV-2 in golden hamsters

A mouse model of SARS-CoV-2 infection and pathogenesis

SARS-CoV-2 entry factors are highly expressed in nasal epithelial cells together with innate immune genes

Rapid reconstruction of SARS-CoV-2 using a synthetic genomics platform

Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro

Retrospective Multicenter Cohort Study Shows Early Interferon Therapy Is

The authors wish to thank all members of the Boston University COVID-19